Survey on Fake Coin and Currency Detection

Transcription

1 376 Survey on Fake Coin and Currency Detection M.Karthika 1, Dr. S. John Peter 2, Mrs. J. Patricia Annie Jebamalar 3 1 M. Phil Scholar, Department of Computer Science, St. Xavier's College, Palayamkottai, Thirunelveli, India 2 Associate Professor, Head of the Department, Department of Computer Science, St. Xavier's College, Thirunelveli, India 3 Associate Professor, Department of Computer Science, St. Xavier's College, Palayamkottai, Thirunelveli, India ABSTRACT This paper makes an attempt to represent the survey on the fake coin and currency detection. Fake notes in the Republic of India are flowing into the system. Now, the identification of fake notes is becoming a very important topic for the researchers. The focus of the currency identification system is on fake currency recognition. The Currency Identity System is important today, it must be accurate. Progressive approach to identifying paper money relies on several steps, together with edge detection, feature extraction, image segmentation, image acquisition, grayscale conversion, and comparison of pictures. This survey paper discusses how to look for fake coins and notes. This paper provides an overview of India's Currency Identification Methods to determine if the malware has been activated. When some pre-training and training techniques are implemented, the validity of the Currency Identity System can be improved. Keywords - Currency circulation, Edge Detection, Fake currency Identification, Feature Extraction, Image Processing 1. INTRODUCTION Modernization of the financial system is the cornerstone of the protection of economic prosperity and the maintenance of social harmony. The Reserve Bank of India is the only fully-fledged bank in India. But some people tend to create counterfeit currency. India's 100, 500, and 2000 counterfeit currency seem to have flooded the system, and there is no meaningful way to deal with them for ordinary people. The common people are the victims of these currencies and 500 are the highest available currency and the maximum fake is done in them. For many years, the currency, sterling silver has spread to society and there is no balance between social harmony and society. Numerous agreements have been made in this regard. Reserve Bank is the only bank that has the sole power to issue bills in India. Reserve banks, like other central banks around the world, occasionally change the design of the bills. In general, anti-fraud measures are included, including good bills, which allow specialists, not to fake. On grinding or parallel-grooved, the edges are used to indicate that no precious metal is scraped. Reserve banks use many fraudulent techniques. Fake or counterfeit cards are the biggest issue of money transactions. For a country like India, it is becoming a major obstacle. Due to the advanced scanning technology, it is easy for people to print fake records using the latest hardware. The detection of fake handwritten records becomes a costly and unsatisfactory process that requires an automatic technique that can process the currency identification process efficiently. 2. LITERATURE SURVEY The effective recognition of the blind banners is used based on the components. This article should mark the daily sign with ultraviolet light. The employer employs the account of the equipment and tries to determine if the ocean's identities, identities and other notes are appropriate for their denomination and authentication. This increases the labor force. On the other hand, if the bank staff uses the system, the results may be correct. Traffic Blur affects system performance, so the true speed of marking marks is reduced. The problem can be summarized as follows: i) Problems with motion blur (ii) Ambiguities that are placed by the imager iii) Technology to download less efficient features. This technique is proposed by Yinglienian [1]. The techniques for detecting fake currency methods based on bit rate reduction techniques. This article introduces a new approach using cost-saving techniques to extract the most important data from the images of fake bills by applying edge detection algorithms. The proposed technique is in the original image distribution, with the gray gradient 256 in its binary image equal 8. This is useful when analyzing the importance of attributes defined by each image of the original image. Higher-bit bits are evaluated for the blue dollar bill image by applying the Canny Edge Finder algorithm. Then the result is compared to real money and other available techniques that are used to detect fake notes. This methodology is proposed by Mohammad H. Alshayeji [2].

2 377 This technique describes an electronic-based blind-based Enhancement System. This paper is designed primarily to support them and make it easier for them to use the currency. Here we offer Android based apps to recognize the currencies from different countries, as well as their denominations for the blind. Image processing techniques such as download features and matching are used to determine the currency. The software runs on the low-end smartphone. We provide voice mail as an input to launch apps and capture images. Then the image is captured and compared to the test image. If the features of both pictures are matched, the audio output is given to the user about the currency name and the country in which it belongs. Otherwise, an error message is provided as a result. This technique is proposed by Nayana Susan Jose and Sherman Siby [3]. This article describes the automatic paper recognition system. This can be useful in banking systems and other commercial areas. This paper describes the recognition of paper currency using digital imaging techniques. Three Indian paper features were chosen to look for counterfeit goods, including identity certificates, security threats and sketches. Download features are made on the image of the currency and are compared with real currency characteristics. The gradient level Sobel operators are used for character extraction. Recognizing accurate paper currency and high speed processes are important for the banking system. This approach has the advantages of simplicity and high speed. This methodology is proposed by Mirza and Nanda [4]. 3. METHODOLOGY 3.1 Artificial Neural Network Artificial neural network Neural Networks, with their remarkable ability to derive meaning from complicated or imprecise data, can be used to extract patterns and detect trends that are too complex to be noticed by either humans or other computer techniques. A trained neural network can be thought of as an "expert" in the category of information it has been given to analyze. This expert can then be used to provide projections given new situations of interest and answer "what if" questions. Other advantages include: Adaptive learning: An ability to learn how to do tasks based on the data given for training or initial experience. Self-Organization: An ANN can create its own organization or representation of the information it receives during learning time. Real Time Operation: ANN computations may be carried out in parallel, and special hardware devices are being designed and manufactured which take advantage of this capability. Fault Tolerance via Redundant Information Coding: Partial destruction of a network leads to the corresponding degradation of performance. However, some network capabilities may be retained even with major network damage. Using Multiple Hidden Layers: Multilayer perceptrons using a back propagation algorithm are the standard algorithm for any supervised learning pattern recognition process and the subject of ongoing research in computational neuroscience and parallel distributed processing. They are useful in research in terms of their ability to solve problems stochastically, which often allows one to get approximate solutions for extremely complex problems like fitness approximation. This method has low computational complexity. This is the big advantage of this approach. But it has one disadvantage that is its [1] accuracy is rather low. 3.2 Local Spatial Feature The fake coin detection method uses the local features of the image for feature extraction. The method is invariant to rotation and translation and also the recognition is being done with the single neutral image as train image. Normalized Local standard deviation filter is used for extracting the spatial intensity changes in the image. The key feature of this approach is that it uses single gallery image per coin for the recognition purpose and produces high recognition accuracy. Using background extraction of the coin and the decomposition of an image into concentric circles increases its invariant property against rotation and translation [2]. 3.3 Hough Transform Hough Transform is usually used in detection programs to detect shapes. Here in the proposed system, we want to apply it to coins of circle shape so the Circular Hough Transform (CHT) is used. The Hough transform is specific to a circle of given radius and the result of the transform gives a likelihood of each pixel being the center of a circle with the given radius. In Circular Hough Transform there exist two spaces, geometric space, and accumulator or (Hough space). Each point in the geometric space on the original circle

3 378 edges generates a circle in the accumulator space. The circles in accumulator space intersect together at (a, b) that correspond to the center (x, y) in the geometric space. Notice that (a, b) is a group of pixels that are likely to be the center of the circle. This method is very fast and has less memory capacity. But the performance of this method [3] is very poor in case of low-resolution images. 3.4 Morphological Operation Approach Morphological recognition algorithm based fake coin recognition capture the image by the camera and converted into a grayscale image for pre-processing. After conversion, image complemented, binary conversion is applied to the image. After conversion, the method for detecting the edge of the cup was performed and passed the discovery in the expansion process. After filtering and expanding, the area is chosen where the maximum number of pennies is and the coins are recognized by the image in the form of the box. Block analysis is used for coins for Indian coin discovery. It provides the clear edges of the coins to improve accuracy for coin detection. This is the great plus of this method. The major drawback of this approach [4] is, this scheme only considers static data files and do not explicitly study the problem of data error localization. 3.5 Rotational Invariant Approach Rotation invariance approach, it is not necessary to place the coin at a specific angle. Image Segmentation helps to reduce the amount of data required for processing. This system takes less time for processing and gives the best results rotation invariance approach, it is not necessary to place the coin at a specific angle. Image Segmentation helps to reduce the amount of data required for processing. This system takes less time for processing and gives the best results. This algorithm using rotation invariance approach it is not necessary to place the coin at a specific angle. Also, it [5] helps to reduce the amount of data required for processing. 3.6 Edge Detection and Enhancement The thresholded (binary) image has irregularities that will be removed with Morphological closing and filling plus area thresholding. The morphologically close image is done by using a disk structuring element to preserve the circular nature of the object. The filling is done to fill in holes in the image that are a set of background pixels that cannot be reached by filling in the background from the edge of the image. Finally, area Thresholding is applied also to remove the small objects which containing fewer than 1000 pixels from the binary image. Recognition time of this method is very less. So this is the advantage of this approach. But it did not use the vertical locations of coin candidates for identifying fake coin because the changes in positioning and respiration between the two views were not considered in this scheme. This approach [6] reduces the accuracy. 3.7 Image Subtraction Having procured both the object and the test image, two subsequent checks are performed narrowing down the recognition process. Coarse Subtraction: The test image is given one full rotation in steps of the fixed angular distance of say 30º. At each instance of rotation, image subtraction is carried between the rotated test image and the input object image. The two images basically are a 2-d array consisting of gray values. Subtracting these array yields gray values begetting the third image. Subtracted Image (r, c) = Object (r, c) test image (r, c) (1) This approach [7] works well with fewer samples and learning result is more robust. But the limitation of this approach is the choice of kernel and extensive memory requirements. 3.8 Harris-Hessian Algorithm Harris-Hessian detector is based on a Harris-Hessian algorithm and it extracted features from the original image of the coin and usually, interest points are used as the feature points. This method is robust to noisy data. This is the merit of this approach. But this approach [8] need to determine the value of the kernel and distancebased learning is not clear. 3.9 Otsu Algorithm The method is done by the user to get the image from the computer or camera used. Once the image is selected, the RGB image of the coin is scanned on both sides. When the selected RGB image is converted to Grayscale. If the image is selected initially, it is gray, so it does not have to be converted. If the image size is small, RGB images can be used for recognition and recognition, and if the size of the image is large, you need to change it. Once you have completed both steps, you need to crop the image to have a pictured image. When the crop is done, the next step is to divide the image into different segments depending on the total amount of pennies in the picture. This is done using a known algorithm known as Otsu Algorithm. This algorithm is a method of pragmatization based on histograms and is known as command on tags. When analyzing the image, it is first divided, then the

4 379 corresponding low method is applied. To obtain overlapping images in one image, morphing operations can be used. The result is high and unaltered radius. Lack of threshold variability sample leads poor performance. This is the disadvantage of this approach [9] Thresholding Thresholding is applied to the input image to separate foreground from the background through hue and saturation values to create a threshold image that could possibly include overlapping or clutter. To apply thresholding [10], first, determine the hue of the background by using the histogram, and then find the first index of the max value of nonzero elements. Second, apply saturation thresholding. After that, determine threshold factor a value that rates the likelihood of being in the background[10] Component-Based Model The component-based model has four main advantages over the global model: 1) the class-specific information is not evenly distributed on the banknote. Some regions cover more obvious class-specific features, while other regions are relatively similar across different classes. It will be more effective to use those more class-specific components in the recognition of banknotes. 2) A component-based model is able to focus on local and stable parts, which very much less than the pattern of an entire banknote under the geometric and photometric changes. 3) Local image features that are generated from components are much less than that from the entire image. This helps to speed up the matching process and reduce memory requirement. 4) A component-based model is more robust in handling partial occlusions. It is empirically impossible to take account of all conditions which cover the spectrum of possible variations that can result from occlusions. In the component-based model, individual components are detected by their corresponding detectors. Partial occlusions only affect the outputs of a portion of component detectors. As long as a certain amount of components are detected, the whole banknote is still able to be recognized. It has low computational complexity. But it [11] is affected by Motion Blur problem. network and weights are initialized to some random values, and the activation function is applied. The output is checked with the targets, if it is not same as targets then it propagates an error back to all previous layers and weights are updated on all layers. Three-layer feed forward Backpropagation Neural Network with sigmoid activation function is trained with the learning parameters. The author concluded that efficient feature extraction technique will improve the accuracy of identification against the forged currency. This approach [12] are not robust for all currency notes Multi Neural Network Multilayer Perception is a new technique to recognize paper currency. The recognition of paper currency has been successfully attempted to be recognized from both sides. It uses two classifications, the Euclidean circle weighing at the right weight and the nervous system. The requested technique is based on the download of certain specifications of the paper currency. In addition to finding and extracting features, the technique also includes various preprocessing steps. Various factors like the image size, edge detection, the Euler number and the correlation coefficient play important role in the recognition process. This method uses two classifiers, the weighted Euclidean. But image quality is produced by this approach [13] is very low and the system does not work properly Negative Correlation Learning The Negative Correlation Learning was to produce different individual NN in the ensemble, so that entire ensemble learns the input pattern completely. ENN is a learning paradigm where a collection of a finite number of neural networks is trained for the same task. The input vectors are applied simultaneously in all the ensembles. The negative correlation learning is to produce the diversity of the individual networks using a penalty term. It can be easily implemented and it can provide high classification performances. The major drawback of this approach [14] is computation complexity. TABLE 1: Comparison of existing methods Methods Description 3.12 Backpropagation Neural Network The Backpropagation algorithm was proposed by Rumelhart, Hinton, and Williams in The Backpropagation algorithm works as a feed-forward neural network but it propagates back the error to previous layers so that weights can be updated in all layers. In the first step, the input is presented to the Artificial Neural Network Local Spatial Features This method is used to separate the fake and real currency. This approach extracts the feature which uniquely identifies the fake currency.

5 380 Heuristic Approach, Hough Transform This technique is used in this paper to find the shapes of the currency. Negatively Correlated Neural Network Ensemble This approach is used to produce different individual weights of fake currency. Morphological Operation Artificial neural network Rotational Invariance Approach Edge detection, Cropping, threshold detection Statistical Approach Image Subtraction Otsu s Algorithm, Hough Transform Component-based framework Radius Basis Function Three-layer backpropagation It is used for each coin separately to detect Indian coins. This method is used to separate the fake and real coin. This approach is used to recognize the coin and note in any direction. This approach is applied to the input image to separate foreground from the background through huge and saturation values to create a thresholded image. This approach is used to drive the statistical feature from the given input image. Image subtraction is carried between the rotated test image and the input object image to find the fake note and coin. It divides the image into various segments depends on the total number of the coins present in an image. This approach is used to speed up the matching process and reduce memory requirement A radial basis function (RBF) is a real-valued function whose value depends only on the distance from the original currency and the fake currency. In this paper, a three-layer backpropagation neural network (BPNN) is employed for fake coin detection by using a concentration based feature construction (CFC) approach. Multilayer Perceptrons Feed-forward Backpropagation Neural Network (BPN) 4. CONCLUSION It extracts the feature value in a multilayer to increase performance. In this approach, the output of the first layer is given as the input of the next layer. This paper discusses various coins and currency identification methods based on image processing. Table 1 compares the work of some researchers, which is useful for short study. Many methods have been proposed till now for the coin and currency recognition, but still, very less work has been done for better accuracy. The comparative experiments show that each of these approaches improves accuracy, smaller feature dimension and faster feature extraction time than the previous approaches. This survey has conducted a study through various literature describing different techniques to determine fake coins and currencies. From this survey, it is clear that the accuracy of the fake coin and currency identification can be improved by applying some efficient pre-processing and feature extraction techniques. REFERENCES [1] Faiz M. Hasanuzzaman, Xiaodong Yang, and YingLiTian, Senior Member, IEEE Robust and Effective Component-based Banknote Recognition for the Blind, IEEE Trans System Cybern C Appl Rev Nov; 42(6): [2] Mohammad H Alshayeji, Mohammad Al-Rousan and Dunya T. Hassoun, Detection Method for Counterfeit Currency Based on Bit-Plane Slicing Technique, International Journal of Multimedia and Ubiquitous Engineering Vol.10, No.11 (2015). [3] Nayana Susan Jose, Sherman Siby, Juby Mathew, Mrudula Das, Android Based Currency Recognition System for Blind, International Journal of Engineering Research in Computer Science and Engineering (IJERCSE) Vol 2, Issue 4, April 2015.

6 381 [4] Rubeena Mirza, Vinti Nanda, Characteristic Extraction Parameters for Genuine Paper Currency Verification Based on Image Processing, IFRSA International Journal of Computing, Volume 2, Issue 2, April [5] Komal Vora, Ami Shah, Jay Mehta, A Review Paper on Currency Recognition System, International Journal of Computer Applications ( ), Volume 115 No. 20, April [6] G. Trupti Pathrabe, Mrs.Swapnili Karmore, A Novel Approach of Embedded System for Indian Paper Currency Recognition, International Journal of Computer Trends and Technology, May to June Issue 2011, ISSN: [7] Pathrabe T, Bawane N.G, Feature Extraction Parameters for Genuine Paper Currency Recognition and Verification, International Journal of Advanced Engineering Sciences and Technologies, Volume 2, 85-89, [8] B.Sai Prasanthi, D. Rajesh Setty, Indian Paper Currency Authentication System using Image processing International Journal of Scientific Research Engineering and Technology (IJSRET), ISSN [9] S. Surya, G. Thailambal, Comparative Study on Currency Recognition System Using Image Processing, International Journal Of Engineering and Computer Science, ISSN: [10] Chinmay Burke, Meghana Sirdeshmukh, M.S.Kanitkar, Currency Recognition Using Image Processing, International Journal of Innovative Research in Computer and Communication Engineering (An ISO 3297: 2007 Certified Organization) Vol. 3, Issue 5, May [11] Dr. Seema Bawa, Shatrughan Modi Dept. of Computer Science and Engineering Thapar University Patiala , India Automated Coin Recognition System using ANN, International Journal of Computer Applications ( )Volume 26 No.4, July [12] Rawan S. Hassoubah (r.hassoubah@gmail.com), Amel F. Aljebry (a_jebr@yahoo.com) Lamiaa A Elrefaei (lamiaabdalah@yahoo.com), Faculties of Computing and Information Technology King Abdulaziz University Jeddah, Saudi Arabia Saudi Riyal Coin Detection and Recognition (proceedings of the 2013 IEEE Second International Conference On Image Information Processing). [13] Velu C M1, P.Vivekanadan2, Kashwan K R3,1 R.S, Department of CSE, Anna University of Technology, Coimbatore , Tamil Nadu, India 2 Director, Knowledge Data Centre, Anna University, Chennai Indian Coin Recognition and Sum Counting System of Image DataMining Using Artificial Neural Networks, International Journal of Advanced Science and Technology Vol 31, June,2011. [14] Unnikrishnan G, Sajith Sethu P (PG Scholar, SCT College of Engineering, Trivandrum, Kerala, India (Assistant Professor, SCT College of Engineering, Trivandrum, Kerala,India Automatic Coin Recognition Using Local Spatial Features IOSR Journal of VLSI and Signal Processing(IOSR-JVSP) Volume 3, Issue 5 (Nov. Dec. 2013), PP e-issn: , p- ISSN No. : www.iosrjournals.org [15] Suchika Malik, Parveen Bajaj, Mukhwinder Kaur ECE &M.M U Sadopur Ambala India, Sample Coin Recognition System using Artificial Neural Network on Static Image Dataset, Volume 4, Issue 1, January 2014 ISSN: X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: [16] C.M.Velu, HOD of CSE, SKR Engineering College, Chennai , India, P. Vivekanandan, Director, Knowledge Data Centre, Anna University, Chennai , India. cmvelu41@gmail.com, vivek@annaunive.ed, Indian Coin Recognition System of Image Segmentation by Heuristic Approach and Hough Transform (HT) Int. J. Open Problems Compt. Math., Vol. 2, No.2, June [17] Malatesh, Anitha G, Indian Coin Matching and Counting Using Edge Detection Technique, International Journal of Innovative Research in Computer and Communication Engineering, Vol. 3, Issue 2, February [18] Hafeez Anwar, Sebastian Zambanini, Martin Kampel, and Klaus Vondrovec Hafeez Anwar, Sebastian Zambanini, Martin Kampel, Klaus Vondrovec, Ancient Coin Classification Using Reverse Motif Recognition: Image-based classification of Roman Republican coins, IEEE Signal Process, Mag. 32(4): (2015). [19] Nikita Shelgikar, Prof. L.M.R.J. Lobo, Indian Coin Recognition with Rotation Invariance using Radial Blur Technique, IJAIEM 2014, 1M.E. (CSE) Department of Computer Science & Engineering, Walchand Institute of Technology, Solapur, India 2Associate Professor in Department of Computer Science & Engineering, Walchand Institute of Technology, Solapur, India. [20] Rathod Prahaladsinh Kanubha, Y.J.Parmar, A Research - Develop An Efficient Algorithm To Recognize, Separate And Count Indian Coin From Image Using Matlab, International Journal of Modern Trends in Engineering and Research, Volume 02, Issue 02, February 2015.